panic! (the 'impossible' happened)

[iampure@…]

After hours of reducing my code, I finally was able to make a small test case. I am essentially stuck if this doesn't get fixed soon.

Output:

ghc-6.7.20070702: panic! (the 'impossible' happened)

(GHC version 6.7.20070702 for i386-unknown-linux):

nameModule $dMonad{v a3hFy}

Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug

[iampure@…]

It should load fine in 6.7, but evaluating a, makes it panic.

[simonpj]

Thanks for the boiled-down case. I'll look into it.

So it's fine in 6.6.1, right? Can you use that meanwhile?

Simon

[simonpj]

I've boiled it down more:

module Foo where

import Control.Monad.State

newtype L m r = L (StateT Int m r)

instance Monad m => Monad (L m)

zork :: (Monad m) => a -> L m ()
zork = undefined

mumble e = do { modify id; zork e }

Compiling this with -dcore-lint shows the error. More anon.

Simon

[iampure@…]

I am sure my requirements don't fit version 6.6.1. This specific code works with 6.6.1, but I reported other bugs, on which 6.6.1 doesn't suffice. I think most of those bugs have been fixed in the version I report this bug for (which is the reason I compiled 6.7), but I am positive my code base doesn't work with 6.6.1. Conclusion: this is still a blocker for me.

What is "More anon."?

[simonpj]

I've fixed this one. Thanks for a helpful report.

Simon

• Fix Trac #1494

A tricky bug to do with the way the implication constraints are solved in TcSimplify?. See Note [Reducing implication constraints].

[iampure@…]

I was pleased to find out you had solved this so quickly, but unfortunately, I cannot build ghc from darcs now.

I did: ./darcs-all pull -a sh boot make and got:

 http://pastebin.ca/604150

[iampure@…]

Forgot trivial step in build process, my bad... building.